Reconciling neuronal representations of schema, abstract task structure, and categorization under cognitive maps in the entorhinal-hippocampal-frontal circuits

Prefrontal and lateral entorhinal neurons co-dependently learn item-outcome rules

The ability to learn novel items depends on brain functions that store information about items classified by their associated meanings and outcomes1-4, but the underlying neural circuit mechanisms of this process remain poorly understood. Here we show that deep layers of the lateral entorhinal cortex (LEC) contain two groups of 'item-outcome neurons': one developing activity for rewarded items during learning, and another for punished items. As mice learned an olfactory item-outcome association, we found that the neuronal population of LEC layers 5/6 (LECL5/6) formed an internal map of pre-learned and novel items, classified into dichotomic rewarded versus punished groups. Neurons in the medial prefrontal cortex (mPFC), which form a bidirectional loop circuit with LECL5/6, developed an equivalent item-outcome rule map during learning. When LECL5/6 neurons were optogenetically inhibited, tangled mPFC representations of novel items failed to split into rewarded versus punished groups, impairing new learning by mice. Conversely, when mPFC neurons were inhibited, LECL5/6 representations of individual items were held completely separate, disrupting both learning and retrieval of associations. These results suggest that LECL5/6 neurons and mPFC neurons co-dependently encode item memory as a map of associated outcome rules.

Abstract representations emerge in human hippocampal neurons during inference

Humans have the remarkable cognitive capacity to rapidly adapt to changing environments. Central to this capacity is the ability to form high-level, abstract representations that take advantage of regularities in the world to support generalization1. However, little is known about how these representations are encoded in populations of neurons, how they emerge through learning and how they relate to behaviour2,3. Here we characterized the representational geometry of populations of neurons (single units) recorded in the hippocampus, amygdala, medial frontal cortex and ventral temporal cortex of neurosurgical patients performing an inferential reasoning task. We found that only the neural representations formed in the hippocampus simultaneously encode several task variables in an abstract, or disentangled, format. This representational geometry is uniquely observed after patients learn to perform inference, and consists of disentangled directly observable and discovered latent task variables. Learning to perform inference by trial and error or through verbal instructions led to the formation of hippocampal representations with similar geometric properties. The observed relation between representational format and inference behaviour suggests that abstract and disentangled representational geometries are important for complex cognition.

Circuit dynamics of the olfactory pathway during olfactory learning

The olfactory system plays crucial roles in perceiving and interacting with their surroundings. Previous studies have deciphered basic odor perceptions, but how information processing in the olfactory system is associated with learning and memory is poorly understood. In this review, we summarize recent studies on the anatomy and functional dynamics of the mouse olfactory learning pathway, focusing on how neuronal circuits in the olfactory bulb (OB) and olfactory cortical areas integrate odor information in learning. We also highlight in vivo evidence for the role of the lateral entorhinal cortex (LEC) in olfactory learning. Altogether, these studies demonstrate that brain regions throughout the olfactory system are critically involved in forming and representing learned knowledge. The role of olfactory areas in learning and memory, and their susceptibility to dysfunction in neurodegenerative diseases, necessitate further research.

Long-term olfactory enrichment promotes non-olfactory cognition, noradrenergic plasticity and remodeling of brain functional connectivity in older mice

Brain functional and structural changes lead to cognitive decline during aging, but a high level of cognitive stimulation during life can improve cognitive performances in the older adults, forming the cognitive reserve. Noradrenaline has been proposed as a molecular link between environmental stimulation and constitution of the cognitive reserve. Taking advantage of the ability of olfactory stimulation to activate noradrenergic neurons of the locus coeruleus, we used repeated olfactory enrichment sessions over the mouse lifespan to enable the cognitive reserve buildup. Mice submitted to olfactory enrichment, whether started in early or late adulthood, displayed improved olfactory discrimination at late ages and interestingly, improved spatial memory and cognitive flexibility. Moreover, olfactory and non-olfactory cognitive performances correlated with increased noradrenergic innervation in the olfactory bulb and dorsal hippocampus. Finally, c-Fos mapping and connectivity analysis revealed task-specific remodeling of functional neural networks in enriched older mice. Long-term olfactory enrichment thus triggers structural noradrenergic plasticity and network remodeling associated with better cognitive aging and thereby forms a promising mouse model of the cognitive reserve buildup.

From synapses to circuits: What mouse models have taught us about how autism spectrum disorder impacts hippocampal function

Autism spectrum disorder (ASD) is a complex neurodevelopmental disorder that impacts a variety of cognitive and behavioral domains. While a genetic component of ASD has been well-established, none of the numerous syndromic genes identified in humans accounts for more than 1% of the clinical patients. Due to this large number of target genes, numerous mouse models of the disorder have been generated. However, the focus on distinct brain circuits, behavioral phenotypes and diverse experimental approaches has made it difficult to synthesize the overwhelming number of model animal studies into concrete throughlines that connect the data across levels of investigation. Here we chose to focus on one circuit, the hippocampus, and one hypothesis, a shift in excitatory/inhibitory balance, to examine, from the level of the tripartite synapse up to the level of in vivo circuit activity, the key commonalities across disparate models that can illustrate a path towards a better mechanistic understanding of ASD's impact on hippocampal circuit function.

Abstract representations emerge in human hippocampal neurons during inference behavior

Humans have the remarkable cognitive capacity to rapidly adapt to changing environments. Central to this capacity is the ability to form high-level, abstract representations that take advantage of regularities in the world to support generalization 1 . However, little is known about how these representations are encoded in populations of neurons, how they emerge through learning, and how they relate to behavior 2,3 . Here we characterized the representational geometry of populations of neurons (single-units) recorded in the hippocampus, amygdala, medial frontal cortex, and ventral temporal cortex of neurosurgical patients who are performing an inferential reasoning task. We find that only the neural representations formed in the hippocampus simultaneously encode multiple task variables in an abstract, or disentangled, format. This representational geometry is uniquely observed after patients learn to perform inference, and consisted of disentangled directly observable and discovered latent task variables. Interestingly, learning to perform inference by trial and error or through verbal instructions led to the formation of hippocampal representations with similar geometric properties. The observed relation between representational format and inference behavior suggests that abstract/disentangled representational geometries are important for complex cognition.

Biologically plausible local synaptic learning rules robustly implement deep supervised learning

In deep neural networks, representational learning in the middle layer is essential for achieving efficient learning. However, the currently prevailing backpropagation learning rules (BP) are not necessarily biologically plausible and cannot be implemented in the brain in their current form. Therefore, to elucidate the learning rules used by the brain, it is critical to establish biologically plausible learning rules for practical memory tasks. For example, learning rules that result in a learning performance worse than that of animals observed in experimental studies may not be computations used in real brains and should be ruled out. Using numerical simulations, we developed biologically plausible learning rules to solve a task that replicates a laboratory experiment where mice learned to predict the correct reward amount. Although the extreme learning machine (ELM) and weight perturbation (WP) learning rules performed worse than the mice, the feedback alignment (FA) rule achieved a performance equal to that of BP. To obtain a more biologically plausible model, we developed a variant of FA, FA_Ex-100%, which implements direct dopamine inputs that provide error signals locally in the layer of focus, as found in the mouse entorhinal cortex. The performance of FA_Ex-100% was comparable to that of conventional BP. Finally, we tested whether FA_Ex-100% was robust against rule perturbations and biologically inevitable noise. FA_Ex-100% worked even when subjected to perturbations, presumably because it could calibrate the correct prediction error (e.g., dopaminergic signals) in the next step as a teaching signal if the perturbation created a deviation. These results suggest that simplified and biologically plausible learning rules, such as FA_Ex-100%, can robustly facilitate deep supervised learning when the error signal, possibly conveyed by dopaminergic neurons, is accurate.